Command-line flags for ML anomaly detection and model persistence.
| Flag | Description | Default |
|---|
--ml-anomaly | Enable ML-based anomaly detection | disabled |
--ml-clusters N | Number of KMeans clusters | 3 |
--ml-compare | Compare ML vs z-score results | disabled |
# Basic ML anomaly detection
renacer -c --ml-anomaly -- cargo build
# Custom cluster count
renacer -c --ml-anomaly --ml-clusters 5 -- ./app
# Compare with z-score
renacer -c --ml-anomaly --ml-compare -- ./app
| Flag | Description | Example |
|---|
--save-model FILE | Save trained model to .apr | --save-model baseline.apr |
--load-model FILE | Load pre-trained model | --load-model baseline.apr |
--baseline FILE | Compare against baseline | --baseline release-1.0.apr |
# Save model after training
renacer -c --ml-anomaly --save-model baseline.apr -- cargo build
# Load existing model (skip training)
renacer -c --ml-anomaly --load-model baseline.apr -- cargo test
# Regression detection
renacer -c --ml-anomaly --baseline baseline.apr -- cargo build
=== ML Anomaly Detection Report ===
Clusters: 3
Silhouette Score: 0.847
Model saved: baseline.apr
- Training samples: 47 syscalls
- Compression: Zstd
- Size: 1.2 KB
=== Regression Analysis ===
Baseline: baseline.apr (v0.6.3, 47 samples)
Current: 52 syscalls
New anomalies not in baseline:
- futex (avg: 1250µs) - REGRESSION
Silhouette change: 0.847 → 0.723 (-14.6%)
| Flag | Description | Default |
|---|
--anomaly-realtime | Real-time z-score monitoring | disabled |
--anomaly-threshold N | Z-score threshold | 2.0 |